Stability of mental health across the pandemic

Our results indicate no significant changes across total mental health difficulties for adolescents aged 10–15 over three months of the pandemic, nor changes for most specific mental health domains. This is an important contribution given the currently mixed picture regarding adolescent mental health in the pandemic, with considerable variation across studies and mental health domains.^{Reference Newlove-Delgado, Russell, Mathews, Cross, Bryant and Gudka8,Reference Mansfield, Santos, Deighton, Hayes, Velikonja and Boehnke9,Reference Knowles, Gayer-Anderson, Turner, Dorn, Lam and Davis11} There are several possible explanations for understanding how our study sits within that wider literature base. First, much existing evidence is limited by a lack of pre-pandemic measures and thus cannot capture the existing time trends in young people’s mental health.^{Reference Cybulski, Ashcroft, Carr, Garg, Chew-Graham and Kapur17} Without capturing those trends, the counterfactual ‘what would the mental health be if the pandemic had not happened?’ will be biased towards better mental health, and the estimated effect of the pandemic will be exaggerated. Indeed, the more varied findings already identified suggesting a more complex picture of mental health impact do tend to be those studies that, as in the current study, have incorporated pre-pandemic data.^{Reference Newlove-Delgado, Russell, Mathews, Cross, Bryant and Gudka8} Second, studies often use non-probability samples, such as those obtained from social media, that are likely to be less representative of the population than probability samples,^{Reference Pierce, McManus, Jessop, John, Hotopf and Ford18} such as the one used here. Even among studies that incorporate data from both before and during the pandemic, issues of sampling and representativeness issues are common.^{Reference Newlove-Delgado, Russell, Mathews, Cross, Bryant and Gudka8} The use of self-report data may also be a consideration, although this too diverges from the literature; interestingly, Newlove et al^{Reference Newlove-Delgado, Russell, Mathews, Cross, Bryant and Gudka8} report that their meta-analysis of self-reported emotional difficulties indicated evidence of deterioration in emotional difficulties, whereas a corresponding meta-analysis of parent report found no such evidence.

A fourth consideration is the specific time points of data collection in this data-set: July 2020, November 2020 and March 2021. In the UK, each of those time points occurred during stages of the pandemic where rates of infection and related restrictions were not at their peak (specifics vary across UK nations and, indeed, regions, but broadly follow a similar trend in these time points).^22–24 In July 2020, the first substantial wave of COVID-19 infections was generally considered to have ended,²³ restrictions had been eased (to varying extents) across much of the UK, and many school-age adolescents were attending school, even if only to finish the academic year after long absences. In November 2020, a second wave had begun, and rates were increasing, but restrictions were still relaxed (comparative to previous restrictions), and most school-age adolescents were attending school (except in Wales, which implemented ‘firebreak’ school closures).²² In March 2021, many school-age adolescents were returning to schools across the UK following closures.^22,24–26 The SDQ measure used asks respondents to reflect on the last six months, and so does cover peak periods; however, literature has highlighted recall bias as prevalent in measures reliant on such recall, particularly among children and adolescents whose episodic thinking is still developing.^{Reference Demkowicz, Ashworth, Mansfield, Stapley, Miles and Hayes27} Given this, it is difficult to know for certain the extent to which our data can be considered a reflection of the overall pandemic period or of these non-peak periods. There is some evidence that adolescent mental health symptoms shift in line with pandemic changes, with heightened reporting during periods of more restrictive lockdown measures.^{Reference Creswell, Shum, Pearcey, Skripkauskaite, Patalay and Waite28} Thus, although we cannot say with certainty, it seems likely that had the participants been surveyed in the peak of COVID-19-related restrictions, their responses may have differed. However, it is still valuable to understand how mental health patterns changed during these periods, where there was still a high level of uncertainty and worry for many. A final consideration is that there may be longer-term effects of the pandemic that are not captured by the current study – for example, there may be children who were very young during the pandemic who were affected by the lack of social contact.

Indications of continuity in pre-pandemic mental health trends raise considerations for how we conceptualise effects of the pandemic on youth mental health. It has been argued that indications of short-term impact on young people (and the wider population) reflect not a worsening of mental health but healthy, normative, emotional reactions to profound disruption to everyday life, and that we must take care not to pathologise this as a crisis for most young people.^{Reference Danese and Smith29,Reference Siddaway30} It may well be that such emotional reactions are short-lived, context-specific or do not, in the longer term, translate to items targeting general mental health symptomatology (of course, our examination of mental health here is not exhaustive and does not, for instance, cover eating disorders). Similarly, there are questions as to whether we have underestimated our capacity for adaptation in the face of crisis; insights from global resilience research have demonstrated the capacity for multilevel adaptation in the face of disaster across individuals, communities and complex systems.^{Reference Masten and Motti-Stefanidi31} Indeed, several studies have highlighted various adaptive processes among adolescents in the context of lockdowns.^{Reference Demkowicz, Ashworth, O’Neill, Hanley and Pert32,Reference Luthar, Ebbert and Kumar33} In short, the pandemic may not have affected adolescent mental health as significantly in respect to widespread mental health harm as initially feared. Indeed, our longitudinal findings would suggest a general resilience among youth, an important finding highlighting how adolescents were able to cope with the implications of the pandemic.

Improvement in conduct problems

Contrary to our hypothesis, significant change over time was identified for conduct problems, which improved slightly; that is, adolescents reported fewer conduct problems in the context of the pandemic than they had previously. The effect size observed here (β = −0.26) is modest but meaningful considering typical effect sizes in this area and the longitudinal nature of data, where effect sizes are often smaller.^{Reference Adachi and Willoughby34} This finding contradicts evidence showing increased conduct problems during the pandemic.^{Reference Raw, Waite, Pearcey, Shum, Patalay and Creswell16} However, such evidence is from the very early stages of the pandemic, prior to or during July 2020, where mental health effects may have been different, and may be affected by frequently observed sampling problems. Though we cannot offer conclusive explanations, there are potential ones. To aid reader interpretation, the subscale items are: I get very angry and often lose my temper; I usually do as I am told; I fight a lot. I can make other people do what I want; I am often accused of lying or cheating; I take things that are not mine from home, school or elsewhere. One consideration may that in completing our surveys, some adolescents were reflecting on time spent at home rather than school, while others were reflecting on school experiences during an unusual period. For those thinking about recent time at home, it is possible that for many this time was less strictly managed, with different rules and routines to the structured time that adolescents typically experience in daily school life and, perhaps, less active supervision. Challenges around disobedience and fighting may therefore be less pronounced in the home, or perhaps only perceived to be less pronounced by an adolescent self-reporting. Conversely, for those at school, class sizes were often considerably smaller than usual (advised to be less than half the size wherever possible³⁵), and educator guidance included an emphasis on flexibility and patience around behaviour.^{Reference Müller and Goldenberg36} It may be that in this atmosphere adolescents were less inclined to such behaviours, or may haveperceived themselves as engaging in less of this behaviour within more relaxed rules or smaller groups. Overall, this finding is particularly pertinent amid ongoing efforts to understand key trends such as growing rates of school suspension and exclusion following pandemic restrictions, and whether these link to behavioural issues or changes in practice (e.g.^{Reference Daniels, Thompson and Emery37}). Ongoing research using multiple methods to examine this outcome across later points in the pandemic and in the context of school closures and re-openings would be valuable.

Homogeneity of results across groups

A surprising finding is the lack of significant group differences across changes to total mental health difficulties associated with pandemic surveys. There was widespread concern that there would be disproportionate pandemic mental health effects, and a number of studies reported greater impact for some groups, such as girls or young women^{Reference Gagné, Schoon, McMunn and Sacker12} and children from more deprived households.^{Reference Hu and Qian38} However, we note that another study did not find any heterogeneity in the responses to the pandemic.^{Reference Waite, Pearcey, Shum, Raw, Patalay and Creswell13} Our findings may also reflect the specificity of impact for particular groups; for instance, evidence does suggest that facets of socioeconomic disadvantage were associated with worsened mental health among adults during the pandemic, but perhaps such challenges had fewer direct implications for mental health among the majority of adolescents.^{Reference O’Connor, Wetherall, Cleare, McClelland, Melson and Niedzwiedz39} There were some groups that we could not examine, because of lack of available data, for whom this might not have been the case – for example, those with prior mental illness, special educational needs or neurodevelopmental disorders – as there is evidence of impact on these groups more widely (e.g.^{Reference Raw, Waite, Pearcey, Shum, Patalay and Creswell16}). This may also differ from some of the characteristics we examined here, such as deprivation, in being more related to one’s health or conditions relative to the wider socioeconomic context.

Strengths and limitations

The current study offers several strengths, including our use of self-reported mental health data, modelling of pre-pandemic trends and sample representativeness in the (weighted) data. However, we also acknowledge various limitations. As explored earlier, the time points of our data map on to less intense periods of the first year of the pandemic, avoiding peaks of infections and the height(s) of restrictions, and tend to be near the beginning and end of substantive school closures or the strictest social-distancing measures. This is a strength in that it allows us to ‘zoom out’ beyond particularly disruptive periods that may prompt more immediate emotional reactivity; however, our lack of data in this sample for such periods limits our ability to understand more subtle shifts across peaks and troughs of the pandemic. Our study offers important, rigorous findings of trends across this period, which should be interpreted carefully in the context of ongoing shifts in discourses linked to the pandemic.

We also highlight measurement issues; though the SDQ has demonstrated acceptable reliability and criterion validity, this measure has been critiqued for issues including readability and age-appropriateness,^{Reference Black, Mansfield and Panayiotou40} and it is also particularly brief with limited response options. Future research would benefit from more comprehensive mental health assessments and continued triangulation across varying measurement approaches. We note that the questionnaire did not include the ‘impact supplement’, which assesses functional impairment from symptoms across an individual’s life. Though this was not problematic, as we were interested in impact on mental health difficulties rather than the functional impairment incurred through such difficulties, building an understanding of this using other data-sets can support understanding of the implications of mental health difficulties, specifically in the COVID-19 context. Finally, the study sample has a substantial proportion of non-response, which was higher during the pandemic than before it. This raises the possibility of bias if the factors influencing non-response are related to the outcome, mental health. To mitigate this, survey weights were calculated based on area-level characteristics of the sample, and the weighted sample characteristics were found to be similar before and after the pandemic. These weights should adequately address the potential bias unless there is an unaccounted-for relationship between non-response and mental health. For instance, if households with single parents were less likely to respond during the pandemic and were also more affected by it, the weighted results may still be biased. This is because single-parent status was not included in the weighting model, meaning any systematic differences related to this characteristic would remain uncorrected.

Overall, the current study offers evidence, drawn from a representative sample, that youth mental health did not deteriorate over time across three months of the pandemic. This finding is at least applicable to those aspects of mental health measured by the SDQ. This is an important contribution to the complex evidence base as we continue to unpack the mental health implications of the pandemic, and it may support the idea that most young people’s emotional responses to the pandemic were represented by normative and short-lived changes, rather than a widespread and ongoing public health crisis. Evidence of a reduction, albeit a modest one, in self-reported conduct problems is of interest, and should prompt further exploration of the ways that behaviour was affected or responded to in the context of the first year of the pandemic, and potential shifts and implications over time (e.g. implications upon returning to school). Our research also highlights the importance of monitoring overall trends, including in representative sampling, and of not ignoring the focus on pre- and, indeed, post-pandemic worsening trends: these must continue to be given dedicated attention. Even prior to the pandemic, the mental health of adolescents was deteriorating;^{Reference Cybulski, Ashcroft, Carr, Garg, Chew-Graham and Kapur17} budget cuts and wider challenges meant that mental health provision, including that offered in schools, was described as being at ‘breaking point’.⁴¹ These examples of an already existing escalation of mental health difficulties situate the current study as a further contribution to an evidence base documenting shifts (or lack thereof) at particular points in time. ‘Crisis’ has served as an ongoing narrative in young people’s mental health for several years, and this was further amplified in relation to the pandemic. As we move further away from major restrictions linked to the pandemic in the UK, a continued measured response by researchers is necessary to establishing the extent of the pandemic’s impact on young people’s mental health and, in turn, appropriate policy and practice contextualisation and responses to rates of difficulties.